Python raises TypeError: forward() got an unexpected keyword argument ‘labels’ error in Pytorch BERT when you pass an argument called ‘labels’ to the forward() method, but the method does not expect this argument.
To fix the TypeError: forward() got an unexpected keyword argument ‘labels’, remove the ‘labels’ argument from the forward() method. If you are using a pre-trained model, you may need to fine-tune the model to accept the ‘labels’ argument.
In PyTorch, the BERT model does not use the ‘labels’ argument in its forward() method.
You can separate the loss computation from the forward pass. First, perform a forward pass to get the logits from the model, and then compute the loss using the appropriate loss function (e.g., CrossEntropyLoss for classification tasks).
import torch from transformers import BertTokenizer, BertForSequenceClassification from torch.nn import CrossEntropyLoss # Loading the model and tokenizer model_name = "bert-base-uncased" tokenizer = BertTokenizer.from_pretrained(model_name) model = BertForSequenceClassification.from_pretrained(model_name) # Prepare your input input_text = "Your input text goes here" tokens = tokenizer(input_text, return_tensors="pt") input_ids = tokens["input_ids"] attention_mask = tokens["attention_mask"] # Forward pass to get the logits logits = model(input_ids, attention_mask=attention_mask).logits # Prepare your labels # Assuming you have a binary classification task labels = torch.tensor().unsqueeze(0) # Compute the loss loss_function = CrossEntropyLoss() loss = loss_function(logits, labels)
Replace the model name and input text with the appropriate values for your use case.
I hope it helps!